Comparison of experiment and the proposed general linear

Jan 1, 1986 - Whole Range of Chain Dynamics in Entangled Polystyrene Melts Revealed from Creep Compliance: Thermorheological Complexity between ...
0 downloads 0 Views 1MB Size
Macromolecules 1986,19, 159-168 kcal/mol is supported by thermal degradation experiments, which are found to proceed in two stages: an initial stage characterized by low U (-25 kcal/mol) and accounting for 1-10% weight loss, and a second stage for the remaining part which degrades with higher U (-55 kcal/mol) (see: Regel, V. R.; Amelin, A. V.; Pozdnyakov, 0. F.; Sanfirova, T. P.; Ioffe, A. F. Znt. Symp. Macromol. 1972, Helsinki, Finland, July 2-7,

1972,5,163).That initial stage of the degradation process has been explained by the existence of weak links in the polymer, which, in turn, determine its mechanical properties. (11) Termonia, Y.; Smith, P. in preparation.

159

(12) Smith, T. L. J. Polym. Sci. 1968,32,99;SPE J. 1960,16,1211. (13) Bueche, F. J. Appl. Phys. 1955,26,1133. (14) Bueche, F.; Halpin, J. C. J. Appl. Phys. 1964, 35, 36. (15) Vincent, P. I. In "Encyclopedia of Polymer Science and Technology"; Wiley: New York, 1967; Vol. 7, p 292. (16) Capaccio, G.; Gibson, A. G.; Ward, I. M. In "Ultra-High Modulus Polymers"; Ciferri, A., Ward, I. M., Eds.; Applied Science: London, 1979; p 58. (17) Barham, P. J.; Keller, A. J. Polym. Sci., Polym. Lett. Ed. 1979, 17, 591.

Comparison of Experiment and the Proposed General Linear Viscoelastic Theory. 2. Stress Relaxation Line Shape Analysis Y.-H.Lin Exxon Chemical Company, Plastics Technology Center, Baytown, Texas 77522. Received March 26, 1985 ABSTRACT For testing the proposed general linear viscoelastic theory, the molecular weight (MW) range of the studied samples is extended to as low as 1.24Me,overlapping the previous MW range from 13Meto 31Me. It is found that the proposed general theory describes quantitatively the linear viscoelastic relaxation spectra over the entire MW range that has been studied. The conclusions about the px(t), p&), and pc(t) processes made in paper 1 are further confirmed. Especially, the theory explains why and how the modulus plateau gradually disappears with decreasing MW. From the spectrum line shape analysis in the p A ( t ) region, a very significant phenomenon is discovered: With decreasing MW, the friction coefficient for the polymer chain segment to move in the direction perpendicular to the chain contour decreases and converges to the friction coefficient value along the chain contour, which is independent of MW. The effect is attributed to the anisotropic distribution of free volume on the polymer chain-i.e., more free volume at the polymer chain ends than in the interior portion of the chain.

I. Introduction The effect of chain entanglement on polymer viscoelastic behavior has been an important research subject of polymer physics.'P2 The concept of chain entanglement started more than half a century ago. Not until the introduction of the idea of reptation by de Gennes3 to account for the constraint effect of entanglement was it possible to describe the dynamics of long polymer chains in concentrated systems in a meaningful way. In 1978, Doi and Edwards"' recognized the important relation of the reptation motion and the mechanical properties of a concentrated system and derived a rheological constitutive equation. Their theory has successfully explained many characteristic features of viscoelasticity of concentrated polymer systems. The Doi-Edwards theory mainly dealt with the polymer dynamics and viscoelastic behavior in the terminal region. In the previous report (referred to as paper 1 below),8 it was theoretically shown how four modes of polymer molecular dynamics affected the linear viscoelasticity of flexible linear polymers. The four dynamic modes p A ( t ) , px(t), p B ( t ) , and p c ( t ) are respectively the Rouse chain motion between two adjacent cross-linked points (assuming entanglement points fixed), the chain slippage through entanglement links, the primitive chain-length fluctuation, and the reptational motion corrected for the chain-length fluctuation effect. A general stress relaxation function containing all these dynamic modes was obtained. It was shown that the theory is universal as we normalize molecular weight (MW) with respect to the entanglement MW, Me, which can be determined from the plateau modulus. The theory was used to analyze the stress relaxation line shapes of three polystyrene samples of very narrow MWD and of MW's ranging from 186OOO to 775 000. Consistent good agreements between theory and experimental results were obtained. The theory also explained the MW dependence of the zero-shear viscosity (qo 0: w.4 for M > M , 0024-9297/86/2219-0159$01.50/0

and

v,, M for M < M,) 0:

and the steady-state compliance

Je.

Experimentally, the stress relaxation modulus does not show a clear plateau region in the medium and low-MW range ( MW > Me),while K is independent of MW. The physical meaning of this phenomenon is explained in terms of the free volume associated with the polymer chain ends.

11. Theory Following a step-shear deformation and before the 0 1986 American Chemical Society

Macromolecules, Vol. 19, No. I , 1986

160 Lin

F40 NBS L11 F10 P7 P5 F4 L3 P3 F2

Mw 422 000 179000 110000 102000 68 000 51 500 43 900 35 100 34 500 16700

Table I M,/M,, 1.05 1.07 1.02 1.01 1.03 1.02

characterizationa TSK NBS LF TSK PS PS TSK LF PS TSK

rAp =

"TSK: Toyo Soda Manufacturing Co., Japan, with light scattering, intrinsic viscosity, and GPC methods. NBS: National Bureau of Standards with light scattering and osmometry. PS: Polysciences, Inc. with GPC. LF: L. J. Fetters, Exxon CRSL, with GPC.

polymer chain has a chance to slip through an entanglement link, the stress relaxation of the polymer behaves like that of a cross-linked rubber. Thus, the Rouse chain motionspg between two cross-linked points (or fixed entanglement links), p A ( t ) , is the main stress relaxation mechanism. As the polymer chain slips through the entanglement links to equilibrate segmental density along the chain contour, the modulus (for a simple shear) relaxes from the value GN', given by 3pRT G N ' = -( (E*u),(E*u),) MeX to the value GN, given by

K73 Me2 24 sin2 (7rP/2Ne) N,2

with

and

with rC = K ( M / M , ) [ l- ( M e / M ) 1 / 2 ] 2

(12)

All the relaxation times in the above equations are proportional to the constant K , given as

K = (b2N2/kTr2w

where E is the deformation tensor for simple shear l h 0 E = O l O

Io

0 11

(3)

Equation 1 is from the theory of rubber elasticity,lOJ1 and eq 2 was obtained by Doi and E d ~ a r d s .It~ has been shown that eq 2 describes the measured damping function in the terminal region very well.12-14 The particular dynamic mode, p X ( t ) ,describing the slippage of the polymer chain through entanglement links is a logical consequence of the shown validity of eq 2. For a simple shear deformation in the linear region, G N is smaller than G N ' by a factor of 4/5. After the p x ( t ) process, the processes that relax the remaining polymer stress are the primitive chain-length fluctuation, pB(t), and the reptational motion corrected for the chain-length fluctuation effect, pc(t). Both these two modes of motions relax the polymer stress by moving the chain end through an entanglement link (i.e., disentanglement). The chain-length fluctuation process relaxes the stress at both tube ends to a length proportional to ( M e / M ) 1 / 2The . remaining tube length proportional to 1 - (Me/M)1/2 is relaxed by the ~ ( process. t ) While the pc(t) process is basically reptational, &(t) is a Rouse-type process. Although chain length fluctuation and reptational processes proceed simultaneously, the theoretical expressions of p&) and r c ( t ) correspond to two decoupled processes.8 In the wide MW range, M > Mc, the pc(t) process is slower than the p&) process. At Mc, the characteristic times of k B ( t ) and p c ( t ) cross over. Including all the four dynamic modes pA(t), p x ( t ) , p B ( t ) , and p&), a general functional form for the stress relaxation modulus after a step-shear deformation in the linear region has been derived as8

(13)

In the high-MW region (>lOMe),it was found from analyzing the experimental results that the K value in eq 6 (denoted as K'below) was a few times greater than the K value in eq 10 and 12. Both the static quantities, B and C, are expressed in terms of the reduced MW, MIM,. As we normalize 7B and rc with respect to 7 A 1 , the relaxation times are functions of the reduced MW as well. Thus, the theory is universal in terms of the reduced MW. The entanglement MW, Me, can be determined independently from the plateau modulus, G N , using the equation

Me = 4pRT/5G,

(14)

By keeping the universality applicable in the p x ( t ) region, we obtained by a combination of a scaling argument and analysis of experimental results TX

= 0.55KMa

(15)

111. Experiment Polystyrene samples of very narrow MWD were used for linear viscoelastic measurements in this study. The details of sample preparation have been described before.I2 The measurements were made with the System Four mechanical spectrometer manufactured by Rheometrics, Inc. Among the polystyrene samples studied, NBS was obtained from the National Bureau of Standards, F40, F10, F4, and F2 from TSK (Toyo Soda Manufacturing Co., Japan), and P7, P5, and P 3 from Polysciences, Inc. L11 and L3 were kindly supplied by L. J. Fetters of Exxon Research and Engineering Co. The methods of characterization and the weight-average MW's (M,) of these samples are listed in Table I. In this study, the measurements were made with a 10 kgcm transducer, which is about five times stiffer than the 2 kgcm transducer used previously. The result of F40 measured with the

Macromolecules, Vol. 19, No. 1, 1986

Stress Relaxation Line Shape Analysis 161

"01 .......

.1

3

10-2

........ ........ ........ I

4

I

100 Tiine

Figure 1. Comparison of the stress relaxation moduli of the F40 and 2 kgcm (-) sample measured with the 10 kgcm transducers. (.-e)

Figure 2. Applied step strain of 0.06 for the stress relaxation measurements. 2 kgcm transducer and reported previously is compared with the one measured with the 10 kg-cm transducer, as shown in Figure 1. After the glassy region (after 2 X lo-' s in Figure l), virtually no difference in line shape is observed between these two results. In this study, we do not attempt to analyze the line shape in the glassy region, which is related to the local segmental motion. Since after the glassy region, the line shape is independent of the stiffness of the transducer, the correction due to the compliance of the transducer should be negligible. This is confirmed by the calculation based on the measured compliance of the transducer. The System Four mechanical spectrometer is capable of performing a step deformation to a strain of 0.06, as shown in Figure 2, which is used in the linear viscoelastic relaxation measurements in this study. The step deformation reaches the fixed plateau s. Thus, for the time region of stress value at about 40 X relaxation of our present interest (i.e., after 0.2 s), the step deformation rising time is sufficiently short. When oscillatory measurements for storage and loss moduli =e made, both the strain signal of the servo system and the stress signal from the transducer are correlated with a standard sine wave to calculate the in-phase and out-of-phase modulus components. Because of these correlationsteps, the obtained modulus values should be more reliable than the stress relaxation measurements. To confirm the measurements and line shape analysis of G ( t ) ,the storage and loss moduli of F10, P7, P5, F4, and F2 were measured.

IV. Stress Relaxation Modulus 1. Experimental Results. The measured stress relaxation curves of different MW's at 127.5 f 0.5 "C are shown in Figure 3. At high M W , a plateau region is clearly

d o4

107

1

106

(sec)

Figure 3. Comparison of the measured stress relaxation moduli of F40, NBS,F10, P7, F4, L3, and F2. noticeable. As the MW decreases, the plateau gradually disappears. The phenomenon has been observed in the measurements of the storage and loss moduli and was well documented in l i t e r a t ~ r e . ' ~ ~ J ~ In the glassy and transition regions, the stress relaxation curves shift to shorter times as the MW decreases. This has to do with the change of the glass transition point, Tg, or free volume in the polymer melt with MW.16 2. Line Shape Analysis. The terminal relaxation time is highly sensitive to MW; it was shown that MWD need to be considered in analyzing the stress relaxation line shape.8J2 In this study, one of our main interests is to analyze the stress relaxation in the MW region where the modulus plateau is greatly reduced. In this MW region, because the relaxation times of the different dynamic modes (Le., p X ( t ) ,p B ( t ) , and p c ( t ) ) are close to each other, it is difficult to do the nonlinear least-squares fitting of the measured line shape to the theoretical form to obtain the MWD of the sample as done previously. Instead, we assume that the very narrow MWD of all the studied samples except NBS and F40 can basically be described by the Schulz MWD1'J8

where

r is the gamma function and (2 + 1)/Z = M,/M,

(17)

Then we convoluted eq 16 with eq 4 in combination with eq 7-12 to calculate the stress relaxation moduli at different 2 values that solely determine the width of the MWD. The optimum 2 value for each sample was determined by the best matching of the calculated and measured stress relaxation curves. The p X ( t ) process is not sensitive to MWD. The contribution of p x ( t ) to the G(t) curve is calculated according to eq 15 by using M , for M . The p A ( t ) process is independent of MW (eq 6). The G(t) curve in the p A ( t ) region is solely determined by the K value in eq 6 (denoted as

K?.

Because of its high modulus value, the glassy region of G(t) is much affected by the compliance of the transducer, as shown in section 111. The agreement between theory and experiment in the glassy region reported previously: whereby the best value of Ne (see eq 6) could be determined, was an error due to negligence of the compliance effect of the transducer. No existing theory is available for analyzing G ( t ) in the glassy region.

Macromolecules, Vol. 19, No. 1, 1986

162 Lin

q P

3 m,L,,d + 10-2

, ,,,,,.1

,

100

,,,,,

~

, ,,,,,,.I

,

,

,,,,r(

1o4

102

,

. ,,J IO'

106

Time (sec)

Tiiiie ( s e c )

Figure 4. Comparison of the measured and calculatedstress relaxation modulus (-) for the F40 sample. Also shown are the separate contributions of the pA(t),p X ( t ) , p&), and k(t)processes. (..e)

-

133

Figure 7. Same as Figure 4 for the F10 sample.

-

Y)

m0

10-2

.-A loo

102

Time

104

[sec)

Figure 8. Same as Figure 4 for the P7 sample. Tiine ( s e c )

Figure 5. Same as Figure 4 for the NBS sample.

5 5

i/ PI

"-Zt

. . . . . . . . i A I . I . I . . . I L Y

10-2

102

loo Tiliie

lo3

(sec)

Figure 9. Same as Figure 4 for the P5 sample. Figure 6. Same as Figure 4 for the L11 sample. In the transition region (the N A ( t ) process region not affected by the glassy relaxation), the calculated G(t)curve is very much independent of the Ne value as long as it is sufficiently large. Ten is a very reasonable number for Ne,193' which we have used for the calculation of the p A ( t ) contribution in this study. In the same way as reported previously,8 three component MWD's are extracted from the stress relaxation curves for F40 and NBS by using a nonlinear least-squares fitting computer program.

The calculated and measured stress relaxation curves are compared in Figures 4-13 for samples F40, NBS, L11, F10, P7, P5, F4, L3, P3, and F2, respectively. The Me value used in all the line shape calculations is 13500, corresponding to GN = 2 X loe dyn/cm2. The three-component MWD's extracted from the G ( t ) line shapes of F40 and NBS through nonlinear least-squares fitting are shown in Figures 14 and 15. The Mw/Mnvalues of the three-components MWD's are 1.16 and 1.08 for F40 and NBS, respectively. The Schulz MWD's used to calculate G(t)'s for the rest of the samples are shown in Figure 16. Their M w / M nvalues range from 1.01 to 1.03, in good agreement

Macromolecules, Vol. 19, No. 1, 1986

Stress Relaxation Line Shape Analysis 163

\ m 3

10'2

o

L

L

102

loo

L

lo4

Time ( s e c )

Figure 10. Same as Figure 4 for the F4 sample.

Figure 13. Same as Figure 4 for the F2 sample.

5 c

p --ml

3

Time ( s e c )

Figure 11. Same as Figure 4 for the L3 sample.

lo4

I o5

M o l e c u l a r Weight

Figure 14. Three-component MWD that is extracted from the stress relaxation modulus of the F40 sample using a nonlinear least-squares fitting computer program (see the text and Figure 4).

Time ( s e c )

Figure 12. Same as Figure 4 for the P3 sample. with the GPC values listed in Table I. In the MW region of L11 to F2, the plateau region of the stress relaxation modulus has virtually disappeared. The line shape of G ( t ) is so lack of a "cornern in the terminal region that if only the reptational process of Doi and Edwards is used to analyze G ( t ) the MWD that can be extracted will be very much broader than it really is. In this MW region, the p B ( t ) process plays a very important role because of the following reasons: First, its relaxation strength, B, (eq 7)becomes of the same order of magnitude as that of the p C ( t ) process (eq 8). Second, the relaxation time T B becomes very close to the TC value. These effects

Figure 15. Same as Figure 14 for the NBS sample (see Figure 5).

are demonstrated by the separate contributions of the p B ( t ) and p c ( t ) processes shown in Figures 6-13. By the inclusion of both the p B ( t ) and p C ( t ) processes in the G ( t ) line shape analysis, the narrow MWD's as shown in Figure 16 can be extracted. The success of the line shape analysis in this medium- and low-MW region, where the modulus plateau is either nondistinct or absent,

Macromolecules, Vol. 19, No. I , 1986

164 Lin

t

i

?5

i~

1

P?

1o5

106

M u i e c u l a r We i gh:

Figure 16. Normalized Schulz MWDs that are used to calculate the stress relaxation moduli of L11 (2 = 60), F10 (2 = 60), P7 (2 = 120),P5 (2 = 120), F4 (2 = 60), L3 (2 = 30), P3 (2 = 50), and F2 (2 = 120) (see Figures 6-13).

t

I-

-

-

- AAA

-

*-r-

-

0

t .”

A’

-4--

- -

-

r - - -1j

a - -

-

Moiecula; Veight

Figure 17. K and K‘IK values as a function of MW that have been obtained from the line shape analysis of the measured G(t)’s. The reproducibility of the K values obtained from G ( t ) measurements on different days is within 20%. To eliminate the effect of temperature drift from day to day, measurements of several different samples were made in the same day. Through overlapping resulta of different days, a small correction was made to the data values. The values shown here are the averages of several such corrected values.

strongly supports the proposed general theory. Furthermore, the K values (see eq 10,12, and 13) obtained for the different samples are independent of MW within experimental error, as shown in Figure 17. This strongly supports that eq 10 and 12 correctly give the MW dependence of rB and rc, respectively. The K values of the F40 and NBS samples are about 30% lower than the average value of the rest of the samples (as indicated by the dashed line in Figure 17). This may be due to the error of the M, value used in the calculation. A 30% error in the K value can be caused by less than 10% error of the M, value (see eq 10 and 12). However, we believe that the accuracy of the M, values of the studied samples is better than within 10%. The 30% lower values of the K constant for F40 and NBS are mainly due to the sensitivity of the G(t)line shape to the MWD width at high MW and the low-MW tail that often exists in the high-MW samples. To simplify the explanation for the effect of low-MW components to G ( t ) , one can assume that the sample

consists of only two MW components. If the MW of the low-MW component is sufficiently low compared to the high-MW component, some of the tube stress associated with the high-MW component will relax in a time shorter than the reptation time of the high-MW component, as the low-MW molecules reptate away (disentangle) from the high-MW molecules (i.e., the tube renewal process). This phenomenon has been observed in our study of the blending law.21 Thus, a small amount of the low-MW component can cause a steeper decline in the observed plateau region and reduce the sharpness of the “corner” in the terminal region. This effectively broadens the MWD that can be extracted from the G ( t ) data. Because the modulus plateau is wider and its At/AG value is larger at higher MW, this effect is enhanced at high MW. The M,/Mn value extracted from the G ( t ) line shape of F40 is significantly increased artifically by the breadth of its MWD and the low-MW tail in the samples (see Figure 23). The large low-MW tail in the NBS sample (see Figure 24) has a similar effect. The M,/Mn values of the three-component MWDs extracted from the G(t) line shape analysis for F40 and NBS are greater than those of the rest of the samples, partly due to its intrinsic broader MWD and partly due to the described broadening effect. The linear additivity assumed in the three-component MWD used to analyze the G ( t ) line shapes of these two samples is not strictly obeyed. This tube-renewal effect caused by the low-MW components is not large and somewhat enhances the fast relaxation processes for F40 and NBS. As a result, the K value extracted from the line shape analysis of G ( t ) is slightly smaller. This explanation is supported by the viscosity data obtained from integration over the measured G ( t )curves of the studied samples as reported in the accompanying paper (paper 3). The obtained viscosity values are in very good agreement with the theoretical curve over the entire MW range. In other words, the viscosities of F40 and NBS do not show a 30% deficiency as their corresponding K values. This seeming paradox can be resolved by the above explanation, assuming that within the MWD range of our current interest (M,/Mn < 1.2), the experimental viscosity value is basically a function of M , only and independent of MWD. An equivalent explanation is as follows: We can approximate the general linear viscoelastic theory by the Doi-Edwards theory and still account for the dominant part of the MW dependence of the zeroshear viscosity. Assuming the linear additivity law applied to the Doi-Edwards theory, GraessleY2showed that vo 0: M&IzM,+, instead of qo 0: Mw3.Since M,,, > M, > M,, in order to make vo 0: MW3 (corresponding to vo 0: Mw3,4 as observed experimentally), the K value, which is obtained from the forced computer fitting to the G ( t ) line shape in terms of the three-component MWD (for the cases of the F40 and NBS samples), has to be smaller than it would be for ideal monodispersity. Very significantly, the K value remains quite independent of MW to the MW value as low as 16 700 ( M , of the F2 sample), as shown in Figure 17. At M , = 16 700, the K value being slightly lower than the average value may be due to slightly larger free volume at this low MW. Overall, the change of free volume or Tgwith MW in the MW region 10Me-Me does not affect the K value but affects the K’ value (the K value of eq 6) greatly. A s explained above, because of the presence of the glassy relaxation process in the short-time region, the relaxation line shape cannot be used to determine the Ne value in the p A ( t ) process. However, the matching of experimental and calculated values in the region of low motional modes of

Stress Relaxation Line Shape Analysis 165

Macromolecules, Vol. 19, No. 1, 1986 -.,

. 10-

.....I

,

4

.

, ruL-

1

.

, '

11.11

..

. I . . .

I

1

.

100

10-2

.,...1

.

""'T

-

m

w

-

7

-

v C

IO2

Frequency ( r a d i s e c )

Figure 18. (a) Comparison of the measured and calculated (-) storage moduli of F10. Also shown are the separate contributions of the p A ( t ) , kx(t), /.&), and &) processes. (b) Comparison of the measured and calculated (-) loss moduli of F10. Also shown are the separate contributions of the PA@), d t ) ,d t ) , and d t ) p~ocesses. (e..)

(.-a)

p A ( t ) is sufficient to determine the K'value. The K'values are obtained from fitting the experimental G(t)values in the transition region to the p A ( t ) process (eq 4-6) shown by its contributions in Figures 4-13. As shown in Figure 17, in the high-MW region, the KYK value is independent of MW and has a value of about 4.2. At MW 140000, it begins to decrease with decreasing MW and reaches a limiting value of 1 a t low MW (MW

-

1

-

Me(+)). It has been proposed8 that for K'/K > 1, the friction

coefficient for the polymer chain segment to move in the direction perpendicular to the chain contour is higher than that along the chain contour. This is a reasonable explanation, if we consider that a certain local extra free volume is associated with the polymer chain end. This extra free volume is always available to the pg(t) and pC(t) processes, in which the polymer chain moves along the chain contour. On the other hand, the portion of this extra free volume available to the p A ( t ) process of a polymer chain depends on the concentration of free chain ends of polymer molecules surrounding it. Thus, the K'value extracted from the p A ( t ) process is MW dependent. Furthermore, the K'/K value approaching the value 1a t low MW supports this explanation. The K'/K values of F40 and NBS become 3.3 after correcting for the 30% underestimation of their K values. The corrected K'/K value in the high-MW region, where

Frequency ( r a d i i e c )

Figure 19. (a) Same as Figure 18a for P7. (b) Same as Figure 18b for P7.

K'/K is basically independent of MW, is indicated by the dashed line in Figure 17. The contributions of the px(t) process to the total G ( t ) line shapes shown in Figures 4-13 are calculated with eq 15 for the MW dependence of its relaxation time 7x. By the combination of a scaling argument and the results of G ( t ) line shape analysis, the form and coefficient of eq 15 were determined in paper 1. The validity of eq 15 is further confirmed by the close fitness of the calculated px(t) contributions to the measured G(t) line shapes of all the samples studied in this report.

V. Storage and Loss Moduli To confirm our line shape analysis of G(t),especially in the MW region where the modulus plateau is either nondistinct or absent, the storage and loss moduli, G'(u) and G"(w),of F10, P7, P5, F4, and F2 are measured and compared with those calculated from eq 4 in combination with eq 5-17. The 2 values used in the Schulz MWD for these samples are the same as those used in the calculation of their G ( t ) curves (Figures 7-10 and 13). For calculating the G'(w) and G"(w) values using the linear viscoelastic relations G'(o) = w J m G ( t ) sin w t dt

(18)

G"(w) = u J m G ( t ) COS ut dt

(19)

a computer program containing the single and coupled

Macromolecules, Vol. 19, No. 1, 1986

166 Lin

v-------

-21 10-2

Figure 20. (a) Same as Figure 18a for P5. (b) Same as Figure 18b for P5.

2

1

.

.

.i,u.l

I

l j l j l

1

, , , ,,,,

I

10-5 10’ Frequency ( r a d l s e c )

1

/,,,

,

.,,.: i?

Figure 21. (a) Same as Figure 18a for F4. (b) Same as Figure 18b for F4.

Fourier transformed terms of the M A ( t ) , p x ( t ) , p ~ ( t )and , up to P = 9 for p A ( t ) and P = 35 for ~ ~ ( t the ) high-frequency region where it occurs (Le., the reciprocal of its relaxation time). On the other hand, the effect and p c ( t ) (see eq 5 , 9, and 11)was used. This is a more of the glassy relaxation process on the G”(w) values extends accurate and efficient way of calculating G’(w) and G”(w) to lower frequencies, because of the mathematical form of than a direct numerical Fourier transform from the caltransforming G ( t ) to G”(o) (eq 19). culated G ( t ) , since the calculation extends over several As M , decreases, the contribution of p C ( t ) to the total decades of frequencies. G’(w) and G”(w) decreases while that of pLg(t) increases. The measured and calculated G l w ) and G”(w) of F10, Simultaneously, the relaxation time pB becomes closer to P7, P5, F4, and F2 are compared in Figures 18-22, reTC and eventually becomes larger than TC at M , M,. spectively. In the comparison, the experimental results Furthermore, a t M , M,, both p B ( t ) and p C ( t ) become are shifted to fit to the theoretical curves calculated with very close to w A ( t ) . All these effects cause the gradual K = 1 X lo4. The K’/K values used for these samples are disappearance of the modulus plateau on G’(w) and the those shown in Figure 17. In each of these G’(w) and G”(w) modulus hump in the terminal region of G”(w) with defigures, the separate contributions of the p A ( t ) , p&), p g ( t ) , creasing MW. and p c ( t ) processes are also shown. Noteworthy is the minor role played by the p c ( t ) process For the storage modulus, G’(w), the agreements between for the F2 sample, whose M , is between M , and Me. In theory and experiment for all the samples are very good the low-MW region, M , > MW > Me,the polymer molecule over the entire frequency range, except the expected difbehaves like an “anisotropic” Rouse-like chain-with the ference in the high-frequency glassy region. In the case Rouse-like motions ( k a ( t ) ) along chain contour direction of the loss modulus, G”(w), good agreements between being different from the Rouse motion ( p ~ ( t )perpendi) theory and experiment are observed up to the modulus cular to the chain contour. value of -1 X lo6 dyn/cm2 for all the samples. Beyond G”(w) 1 X lo6 dyn/cm2, the measured G”(w) values rise VI. Discussion and Conclusion faster than the theoretical curves and cross the measured G’(w) values a t 3-5 X lo6 dyn/cm2, while the theoretical In paper 1, a general linear viscoelastic theory was G’(w)and G”(w) curves do not cross in this high-frequency proposed and three nearly monodisperse polystyrene samples with MW’s ranging from 186000 to 775 000 were or high-modulus region. The difference of G”(w) values between theory and experiment in this region is due to the studied for comparison with the general theory. In this glassy relaxation process, which is not included in the report, the MW range is extended as low as 16 700, overgeneral stress relaxation form (eq 4). The effect of the lapping the previous MW range from 179000 to 422000. It is found that the proposed general theory is applicable glassy relaxation process on the G’(w) values is limited in fic(t)processes

-

-

-

-

Macromolecules, Vol. 19, No. 1 , 1986

Stress Relaxation Line Shape Analysis 167

10.70

20.40

21.10

2i.80 22.50 i l u t i o n Titile

23.20

23.w

24.60

a.302 8 . 0 0

(Minbte5)

Figure 23. GPC spectrum of the F40 sample.

20 70

-

22

io

22 BO 23 M) 24 20 24 Elution Time (Minutes)

w

21 BO

28 30 27 00

Figure 24. GPC spectrum of the NBS sample.

CI

0

21 40

'''...,.I

" j " "

'

'

'

.

.

.

.

I

'

'"....,'

'

10-1 101 Frequency (rad/sec)

.'.....'

'

" "

lo3

Figure 22. (a) Same as Figure 18a for F2. (b) Same as Figure 18b for F2. over the entire MW range that has been studied. While the rates of stress relaxations of the samples studied in this report range over five decades (see Figure 3), the K values extracted from the stress relaxation line shape analysis is independent of MW. This is a very significant result, which strongly supports that the separation of G ( t ) in the long-time region into the p B ( t ) and pc(t) processes, as described by eq 9 and 11, respectively, is valid and that eq 10 and 12 have the correct form for the MW dependence of TB and TC. The constraint effect due to chain entangelemenh as described by the proposed general theory persists to the low-MW regions M, > MW > Me. In the MW region where the modulus plateau on G ( t ) and G'(o) and the modulus hump on G"(o) are either nondistinct or absent, the line shapes of the linear viscoelastic spectra are well described by the convolution of the general stress relaxation form with the Schulz MWD. The 2 value, which determines the width of the MWD, is the only adjustable parameter. The Mw/Mn values extracted from the line shape analysis are in good agreement with those of GPC. The line shape analysis of G(t),G'(w), and G"(w) shows the increasingly important role played by p&) relative to ~ ( as t )the M W decreases. Eventually both p&) and p&) become very close to p A ( t ) . These effects cause the modulus plateau to disappear gradually with decreasing MW. The increasing importance of p B ( t ) at low MW and eq 12 (corrected for the chain-length fluctuation effect) are the major factors that cause the difference between the observed MW dependence of viscosity (vo c1: M3.4) and the

prediction of the reptation chain model

(vo M),which 0:

will be analyzed in more detail in the accompanying paper.

The general linear viscoelastic theory was developed by assuming that the constraint release effect is negligible for ideal monodispersity or extremely narrow MWD. The good agreement between theory and experiment over the entire MW range studied confirms the validity of this assumption. However, we have observed some effect caused by the constraint release due to the low-MW components in the F40 and NBS samples. This effect causes about 30% deficiency of the K values obtained from the G ( t ) line shape analysis for the F40 and NBS samples. The pX(t) process implicitly exists in the Doi-Edwards theory. The MW dependence of its relaxation time is given by eq 15. For all the studied samples, the calculated p x ( t ) contributions closely fit the measured G ( t ) line shapes. These quantitative agreements further confirm the form and the coefficient of eq 15, which had been determined in paper 1. From the data analysis in the p A ( t ) region, a very interesting and significant phenomenon, that the K' value decreases with decreasing MW in the MW region MW < lOM,, is discovered. An explanation for this observed phenomenon is given. Consistent with the explanation is the fact that Tg (related to free volume and friction coefficient) is slightly affected by the decrease of MW in the same MW region. Since K' is the only physical quantity related to friction coefficient that changes with MW, it is strongly suggested that the change of K'with MW is directly related to the change of Tgor free volume with MW. In the accompanying paper, how this effect is related to the Tgcorrection for viscosity will be analyzed.

Acknowledgment. The author thanks Dr. L. J. Fetters of Exxon Research and Engineering Co. for providing two of the studied samples.

168

Macromolecules 1986,19, 168-173

Appendix

GPC spectra Of the F40 and ms The are shown in Figures 23 and 24,respectively. The F40 sample has a width at half-height slightly greater than that of NBS, and both samples have a low-MW tail. See the text for their effects on the G ( t ) line shape.

(18) Tung, L. H. “Polymer Fractionation”; Cantow, M. J. R., Ed.; Academic: New York, 1967. (19) Akcasu, A. Z.; Han, C. C . Macromolecules 1979,12,276. (20) Osaki, K.; Schrag, J. L. Polym. J. (Tokyo) i971,2,541. (21) Lin, Y.-H., unpublished results. The blending law can be basically described by the following stress relaxation function for a blend system consisting of two monodisperse components which are far apart:

References and Notes Ferry, J. D. “Viscoelastic Properties of Polymers”, 3rd ed.; Wiley: New York, 1980. Graessley, W. W. Adu. Polym. Sci. 1974,16, 1. de Gennes, P.-G. J. Chem. Phys. 1971,55,572. Doi, M.; Edwards, S. F. J.Chem. SOC.,Faraday Trans. 2 1978, 74, 1789. Doi, M.; Edwards, S. F. J. Chem. SOC.,Faraday Trans. 2 1978, 74, 1802. Doi, M.; Edwards, S. F. J. Chem. SOC.,Faraday Trans. 2 1978, 74, 1818. Doi, M.; Edwards, S. F.J. Chem. SOC.,Faraday Trans. 2 1979, 75, 38. Lin. Y.-H. Macrocmolecules 1984. 17. 2846. Mooney, M. J. Polym. Sci. 1959,34, 599. Treloar. L. R. G. “The Phvsics of Rubber Elasticitv“. _ . 2nd ed.: Oxford ‘University Press: -London, 1958. Mark, J. E. J. Am. Chem. SOC.1970,92,7252. Lin, Y.-H. J. Rheol. 1984,28, 1. Osaki, K.; Kurate, M. Macromolecules 1980,13,671. Vrentas, C.M.; Graessley, W. W. J . Rheol. 1982,26,359. Onogi, S.;Masuda, T.; Kitgawa, K. Macromolecules 1970,3, 109. Boyer, R. F. Rubber Chem. Technol. 1963,36,1303. Schulz, G. V. 2.Physik. Chem., Abst. B 1939,43, 25.

where Wl and Wz are the weight fractions of the low- and high-MW components, respectively, and T ( t ) is a decaying function ( for the tube-renewal process) with a characteristic time TT which is larger than the value by about an order of magnitude. From analyzing the stress relaxation line shapes of nearly monodisperse MWD samples and the blend systems consisting of two components of such narrow MWD, it is s u g gested that the tube decaying function T ( t )for ideal monodispersity should be flat for a period 7T >> TC and then decline quickly; i.e., the tube renewal process is negligible for ideal monodispersity. In a system where the high MW is 775000 (the F80 polystyrene sample of TSK) and the low MW is 179000 (the NBS sample), the 7c2value is reduced to 0 . 8 5 ~ ~ 2 (rc2 is the iC2 for Wl = 0) at W , = 0.25, 0 . 7 ~ at ~ 2Wl = 0.5, and 0.57~2at Wl = 0.75. With decreasing MW of the low-MW component, the effect of W , on the reduction of 7c2increases. If the low MW is greater than -400000 estimated from extrapolation, ic2 of F80 should not be reduced at W, = 0.5 (i.e., 7c2 = 1~2). The physical meaning of the above equation for a blend system will be explained in detail in a future publication. (‘22) Graessley, W. W. J.Polym. Sci., Polym. Phys. Ed. 1980,18,27.

Comparison of Experiment and the Proposed General Linear Viscoelastic Theory. 3. Zero-Shear Viscosity and Steady-State Compliance Y.-H. Lin E x x o n Chemical Company, Plastics Technology Center, Baytown, Texas 77522. Received March 26, 1985

ABSTRACT It was shown that the proposed general linear viscoelastic theory explained the molecular weight (MW) dependence of the zereshear viscosity (qo a M3“ above M, qo = M below M J , the steady-state compliance, and the transition points M,and Ml. T o do so, some information extracted from the short-time viscoelastic relaxation data in the MW region much higher than M , was used. The reason the calculated and measured MW dependences of q0 in the low-MW region agree so well was not evident. On the basis of the new insights obtained from the line shape analysis of the stress relaxation moduli reported in the accompanying report (paper 2), an explanation is given. Furthermore, it is shown that the proposed general linear viscoelastic theory is applicable above Me, while the Rouse theory is applicable just below Me. The transition point appears rather sharp. The result suggests that as far as the free volume distribution on the polymer chain is concerned, there is a “phase” transition point at Me(+) from “isotropic” to “anisotropic”.

I. Introduction Chain entanglement plays a very i m p o r t a n t role in the viscoelastic properties of concentrated polymer systems. One well-known example is the observed molecular weight ( M W ) dependence of the zero-shear viscosity qo a above the critical molecular weight (MW), M,.1*2 Empirically, the value of M , is about twice the entanglement MW, Me, which c a n be determined from the plateau modulus. Below M , the viscosity increases linearly with M , as predicted by the Rouse theory. Similarly, the measured steady-state compliance Je increases linearly with MW below M , and is i n d e p e n d e n t of MW above M;.lv2

The reptation m ~ d e P predicted -~ that the zero-shear viscosity scaled with MW as qo 0: iW.In paper 1: a general linear viscoelastic theory was developed f r o m modifying the Doi-Edwards t h e ~ r y . “ ~ It was shown that the general theory explained the MW dependence of v0 and J, and the transition points M, and M l . To do so, some information extracted from the short-time viscoelastic relaxation data i n the MW region much higher than M , was used. The reason the calculated and measured MW dependence of qo i n the low-MW region agree so well was not evident. In paper 2: the viscoelastic relaxation spectra of narrow MWD polystyrene samples over a wide range of MW

0024-9297/86/2219-0168$01.50/0 0 1986 American Chemical Society